Brain Research 943 (2002) 38–47www.elsevier.com/ locate /bres

Research report

N uclei and subnuclei gene expression profiling in mammalian brain*Pascal Bonaventure , Hongqing Guo, Bin Tian, Xuejun Liu, Anton Bittner, Barbara Roland,1 1Ranelle Salunga , Xiao-Jun Ma , Fredrik Kamme, Bernhard Meurers, Margot Bakker,

1´Mirek Jurzak, Josee E. Leysen, Mark G. ErlanderJohnson & Johnson Pharmaceutical Research & Development LLC, 3210 Merryfield Row, San Diego, CA 92121, USA

Accepted 20 February 2002

Abstract

Information on the neuroanatomical expression of a given gene is critical to understanding its function in the central nervous system.The integration of laser capture microdissection (LCM), T7-based RNA amplification and cDNA microarrays allows for this informationto be simultaneously generated for thousands of genes. To validate this integrative approach, we catalogued the gene expression profilesof seven rat brain nuclei or subnuclei. A hundred cells from the following seven brain nuclei were analyzed: locus coeruleus (LC), dorsalraphe nucleus (DR), parvocellular division (PA) and magnocellular division (MG) of the hypothalamic paraventricular nucleus (PVN) andCA1, CA3 and dentate gyrus (DG) divisions of the hippocampal formation. Of the 2145 genes investigated, 1402 genes (65%) gave ahybridization signal statistically different from the background level that was defined by non-specific hybridizations to 15 different plantgenes. Validation of our microarray data on four arbitrarily selected genes was confirmed by Real-Time PCR. Previous research showingexpression patterns of ‘signature’ genes (n517) for specific brain nuclei are consistent with our findings. For example, as previouslyshown, enriched mRNA expression encoding the serotonin transporter or tyrosine hydroxylase was found in DR and LC cells,respectively. Interestingly, expression of the serotonin 5-HT receptor mRNA was also found in DR cells. We confirmed this new finding2B

by in-situ hybridization. The hierarchical clustering analysis of gene expression shows that the two divisions of the PVN (PA and MG) areclosely related to each other, as well as the three regions of the hippocampal formation (CA1, CA3 and DG), which also showed similargene expression profiles. This study demonstrates the importance, feasibility and utility of cellular brain nuclei profiling. 2002Elsevier Science B.V. All rights reserved.

Theme: Cellular and molecular biology

Topic: Gene structure and function: general

Keywords: Laser capture microdissection; cDNA microarray; Locus coeruleus; Dorsal raphe nucleus; Hypothalamic paraventricular nucleus; Hippocampalformation; 5-HT receptor2B

1 . Introduction brain nuclei function [2,11,21,27]. However, studying geneexpression patterns in the central nervous system using

Simultaneous monitoring of thousands of genes in DNA microarrays is complicated because the existence ofparallel, using DNA microarrays, provides a powerful and many different neuronal subtypes results in intricate ana-efficient means of understanding the function of previously tomical and functional heterogeneity within the mam-uncharacterized neuronal phenotypes in the brain. Doing malian brain [6,26].so allows one to form new hypotheses regarding the This neuronal heterogeneity consists of anatomicallyunderlying molecular mechanisms by which many given defined nuclei, subnuclei, neuronal clusters and neuronal

subtypes within cortical layers. We have previously dem-onstrated that the integration of laser capture microdissec-

*Corresponding author. Tel.: 11-858-784-3078; fax: 11-858-450- tion (LCM), T7-based RNA amplification and cDNA2040.

microarrays allows us to profile gene expression at theE-mail address: pbonave1@prdus.jnj.com (P. Bonaventure).1 cellular level [13,19].Present address: Arcturus Applied Genomics, 2715 Loker Ave. West,Carlsbad, CA 92008, USA. To determine whether this approach is valid in the

0006-8993/02/$ – see front matter 2002 Elsevier Science B.V. All rights reserved.PI I : S0006-8993( 02 )02504-0

P. Bonaventure et al. / Brain Research 943 (2002) 38 –47 39

mammalian brain, we report results from a feasibility studyfor the gene expression mapping of seven laser-capturedrat brain nuclei and subnuclei. One hundred cells werelaser-captured and gene expression profiled from each ofthe following seven brain nuclei: locus coeruleus (LC),dorsal raphe nucleus (DR), parvocellular division (PA) andmagnocellular division (MG) of the paraventricular nu-cleus (PVN) and CA1, CA3 and DG divisions of thehippocampal formation. Subsequent hierarchical clusteranalysis demonstrated that each of the seven nuclei had aunique gene expression profile.

2 . Materials and methods

2 .1. Tissue preparation

Adult male Wistar rats (423622 g) that were initiallyused as control animals in a behavioral paradigm weresubsequently decapitated and their brains were removed,placed in cryomolds, covered with Tissue-Tek tissueembedding medium (OTC) and snap-frozen in dry ice-

Fig. 1. Laser capture microdissection from Nissl-stained sections (10cooled 2 methyl butane (260 8C). Ten-micrometer-thick mm). (a) Hippocampal formation including CA1, CA3 and DG beforecoronal sections in the mid-regions of the PVN (bregma— capture. (b) Hippocampal formation after capture of CA1 cells. (c) CA1

captured cells. Scale bar represents 150 mm.1.80 mm), hippocampal formation (bregma –3.30 mm),DR (bregma—1.78 mm) and LC (bregma—9.80 mm) werecut using a cryostat, mounted on non-coated, clear micro-scope slides and immediately frozen on a block of dry ice. by brief centrifugation. The samples were then processedThe sections were stored at 270 8C. for two rounds of T7-based RNA amplification as de-

All efforts were made to minimize animal suffering and scribed [19] with modifications for the DNA and RNAreduce the number of animals used. purification steps. The Quiaquick PCR purification kit

(Qiagen, Valencia, CA) and RNeasy Mini Kit (Qiagen,2 .2. Laser capture microdissection Valencia, CA) were used for purification steps. Polyinosine

(100 ng) was added to each sample before purification ofA quick Nissl (cresyl violet acetate) staining was used to DNA or aRNA.

identify the neurons as described by Salunga et al. [19].The PixCell II System from Arcturus Engineering, Inc. 2 .4. Labeling method(Mountain View, CA) was used for LCM. One hundredneurons were captured in each brain region of interest in Random hexamer primer was added to 10 mg aRNA,each animal (LC, DR, MG, PA, CA1, CA3 and DG; see denatured at 70 8C for 10 min, then transferred to 4 8C. AFig. 1 for an example of laser-captured cells in the CA1 cocktail containing SuperscriptII reverse transcriptase, 25region). mM dA,T,GTP, 1 mM dCTP, 1 mM Cy3-dCTP (all from

Amersham Pharmacia Biotech, Uppsala, Sweden), RNase2 .3. RNA extraction and amplification of LCM samples Block (Stratagene, San Diego, CA) and 53 Superscript

buffer was then added. Reactions were incubated at 25 8CThe total RNA was extracted from the individual LCM for 10 min, followed by an incubation at 37 8C for 2 h.

samples [19] with modifications as specified below. For Template aRNA was removed using RnaseA, and labeledtotal RNA extraction, a small volume (20 ml /LCM cDNA was purified using the Qiaquick PCR purificationsample) of denaturing buffer (Micro RNA isolation kit, kit.Stratagene, San Diego, CA) containing 2-mercaptoethanol(10 ml /ml, Sigma, St Louis, MO) and polyinosine (300 2 .5. cDNA microarrayng/LCM sample, Sigma, St Louis, MO) was used. Thesamples were incubated at 42 8C for 10 min and loaded A cDNA microarray containing 2145 cDNA clones wasonto a pre-rinsed Microcon-100 column (Millipore, Bed- used in this study. cDNA clones were obtained commer-ford, MA). After two rinses with 500 ml Rnase-free water, cially from Research Genetics (IMAGE consortium),the columns were inverted and the samples were collected Incyte Genomics, as well as internal sources. All clones

40 P. Bonaventure et al. / Brain Research 943 (2002) 38 –47

were verified by DNA sequencing. Each clone was printed internal standard for normalization: 59 primer: TGAGCA-as two independent spots on a given chip, left and right CTGGGGAGAAAGGATTGG, 39primer: TCGGAGATG-side, respectively. Two chips were used for each sample. GTGATCTTCTTGCTGG.Every gene in each sample, therefore, has four data points. The 2nd round amplified RNA of each nucleus fromA contact pin microarrayer (Generation III Array Spotter, three independent captures were pooled and reverse-tran-Molecular Dynamics) was used to spot the clones in scribed into cDNA using random hexamers and theduplicate. Microarrays were hybridized and scanned with a Superscript Pre-amplification kit (Gibco). The reactionconfocal laser scanner (Array Scanner, Molecular Dy- after reverse-transcription (20 ml) was diluted to 80 ml bynamics) with excitation wavelength set at 532 nm, collec- adding 60 ml water. Two microliters of the diluted cDNAtion filter set through a 550- to 600-nm band width filter were used for Real-Time PCR. Real-Time PCR reactionsand the PMT set to 630 V. Image analysis was carried out were set up using LightCycler-DNA Master SYBR Green Ias previously described [19]. kit (Roche) following the manufacturer’s instruction. The

PCR reaction was performed for 40 cycles using a2 .6. Statistical analysis SmartCycler machine (Cepheid) as follows: 1 s at 94 8C,

10 s at appropriate annealing temperature (Trp6A: 62 8C;Since the left and right sides of a chip are mere Clcn2: 62 8C; NF-L: 60 8C; Gabt1: 62 8C; cyclophilin:

duplications and each side has a full collection of cDNA 64 8C), and 15 s at 72 8C with 1 min of 94 8C treatmentclones, data from each side of the chip were treated as one before starting thermal cycles.set. Each data set was multiplicatively normalized to the75th percentile value, which was set to value 100%. Data

2 .8. In-situ hybridization histochemistrywere then normalized using a smooth-spline function inS-Plus software (Insightful) to further adjust data points at

In-situ hybridization histochemistry was performed ac-extreme ends. Fifteen different plant cDNAs were used as

cording to the procedures described previously [1] onnegative controls on the chip. The 75th percentile value of

tissue sections adjacent to that used for LCM. The plasmidall plant data points in all experiments (value 60.43) was

(pINCY) containing the cDNA coding for 5-HT receptor2Btreated as ‘background’. The intensities of each gene, 12 35(GB: X66842) was linearized and [ S]UTP-labeled cRNAdata points each (three rats, four data points each), were

probes were generated by in vitro transcription with T7compared to the ‘background’ using one-sample one-sided

polymerase (sense probe) or SP6 polymerase (antisenset-test in S-Plus. Genes are considered ‘expressed’ if the 6probe). Sections were hybridized at 50 8C with 1310difference to the ‘background’ is significant (P-values# 35c.p.m. of [ S]UTP-labeled cRNA probe (80 ml per0.05). The median value of all 12 data points for each gene

section). After high stringency washes (65 8C), the sectionswas used to represent the intensity of that gene, and the

were exposed to a Fujifilm Imaging Plate (BAS-SR2025)95% confidence limits for the median based on the

for 5 days. The Phosphor Imaging Plate was scanned usingfunction of the quantiles were calculated using S-Plus and

a Fuji Bio-Imaging Analyzer System. The autoradiogramsalso presented. Logarithmical-treated data (log 2) of each

generated by the scanner were visualized using Im-nucleus were used for both correlation and hierarchical

ageGauge V3.12 software (Fujifilm).clustering. Average linkage clustering with centered corre-lation (Cluster, Stanford) was used to compare the simi-larity among the samples. The results are presented usingTreeView (Stanford). 3 . Results

2 .7. Real-Time PCR 3 .1. Gene expression profiling of brain nuclei isreproducible

21Primer preparation: Rat transient receptor potential Cachannel 6A (Trp6A, GB: AB051212): 59 primer: Gene expression profiles were measured from sevenGTGCCAAGTCCAAAGTCC, 39primer: TCGTTCAC- different laser-captured brain nuclei or subnuclei in threeTTCATCACTCTCC; rat chloride channel 2 (Clcn2, adult rats. The seven gene expression profiles were gener-GB:NM 017137): 59 primer: CAGCAGTGACACAGAG- ated by capturing approximately 100 cells from eachACC, 39 primer: GGGAGGTGTTGAAAATGG; rat mi- nucleus from a given animal. To estimate the repro-crofilament, light polypeptide (NF-L, GB:NM 031783): 59 ducibility of independent captures, RNA amplificationsprimer: AAGGATGAGTCTGAAGATGC, 39 primer: AG- and microarray hybridizations, pair-wise similarities ofGATCTGGAACTCAACTGG; rat GABA transporter pro- gene expression profiles were compared by calculating atein (Gabt1, NM 024371, annotated as Sequence 18 from linear correlation coefficient (Fig. 2). Linear correlationsPatent EP 0135277 on the array): 59 primer: TGTAGCAG- coefficients (R) among the three rats ranged from 0.93 toGGTCGTGTGG, 39 primer: CACAGTTCTCACAGTAA- 0.99 for all samples except LC (Fig. 2a). Lower correlationTTGTACG. Rat cyclophilin (GB: AA818858) was used as coefficients (0.83–0.91) were found in LC (Fig. 2a). These

P. Bonaventure et al. / Brain Research 943 (2002) 38 –47 41

Fig. 2. Experimental reproducibility of LCM, T7-based aRNA amplification and cDNA microarray procedure. Pair-wise similarities of gene expressionprofiles measured by the linear correlation coefficient. Logarithmically (log 2) treated data corresponding to 2145 genes were used in each case. (a)Pair-wise linear correlation coefficients (R) among three rats used for each brain nucleus. (b) Pair-wise scatter plots showing the correlation betweendifferent nuclei. The linear correlation coefficients are indicated in the lower right corner of each scatter plot. Histograms indicating the distributions of thedata were also shown as bar graph. Plots and calculations were made using S-Plus.

data suggest that independent capture of 100 cells from a correlation between different nuclei. Correlation coeffi-given nucleus / subnucleus yielded reasonable reproducibil- cients between the different nuclei investigated rangedity of gene expression profiling. from 0.67 to 0.83 (Fig. 2b). Within the PVN, a high

Pair-wise scatter plots were also used to determine the correlation coefficient was observed between the PA and

42 P. Bonaventure et al. / Brain Research 943 (2002) 38 –47

Table 1 the PVN, the expression level of dynorphin mRNA wasNumber of genes giving a hybridization signal statistically different from higher in MG compared to PA as previously shown by ISHthe background level (defined as ‘expressed’, see text for details) in each

[24]. Other striking examples include ‘enriched’ expres-laser-captured nucleus (100 cells). The number of genes that are 1.5, 1.75sion of mRNAs encoding tyrosine hydroxylase (GB:or 2 fold ‘enriched’ in one nucleus with respect to all the six other nuclei

are indicated. A total of 2145 genes were interrogated M10244) and serotonin transporter (GB: AA819178) inLC and DR, respectively [23,24] (Table 2). Low levels of‘Expressed’ Fold enrichmenttyrosine hydroxylase mRNA were also detected in PVN

.1.5 .1.75 .2and DR as described in the literature [8,25,29]. The

CA1 898 21 6 3 restricted expression of mGluR1 mRNA (GB: M36418) toCA3 847 40 26 18 the hippocampus was in agreement with the literature [22].DG 795 26 12 7

However, in contrast to our findings, higher expressionDR 997 49 23 10levels in CA3 and DG compared to CA1 were reported inLC 742 85 67 46

MG 801 19 7 3 an ISH study by Shigemoto and colleagues [22]. FurtherPA 785 14 5 2 examples of gene expression profile corroborated by others

are given in Table 2 (a total of 17 genes is listed in Table2).

MG division (0.92). Within the hippocampal formation(CA1, CA3, DG), the correlation coefficients ranged from

3 .4. Real-Time PCR verification of cDNA microarray0.80 to 0.89 (Fig. 2b).

data

3 .2. The signal intensity of 65% of the analyzed genesUsing Real-Time PCR, four genes were arbitrarily

was statistically different from the background levelselected to validate our microarray data: NF-L (GB:M25638); Trp6A (GB: AB051212); Gabt1 (GB: I07880)

Of the 2145 genes analyzed, 1402 (65%) gave aand Clcn2 (GB: NM 017137). The housekeeping gene

hybridization signal statistically different from the back-cyclophilin (GB: AA818858) was used to normalize both

ground level. These genes were defined as ‘expressed’ (seecDNA microarray and Real-Time PCR data. The results

Material and methods section for details). The number ofpresented in Fig. 3 show that the expression patterns

genes ‘expressed’ within one cell type ranged from 742 toobtained with Real-Time PCR was consistent with the

997 (Table 1).cDNA microarray results. Higher sensitivity was observed

A search for nucleus-specific ‘enriched’ genes waswith Real-Time PCR. For example, according to the cDNA

carried out by determining genes whose value in amicroarray data, the pattern of expression of Trp6A was

particular cell type was at least 1.5-, 1.75- or 2-fold higherrestricted to DG and CA3 whereas Real-Time PCR showed

than in any of the remaining six. The number of genesexpression of Trp6A in all the nuclei.

‘enriched’ in one of the seven investigated brain nucleiranged from 14 to 85 using the 1.5-fold threshold (Table1). As expected, by increasing the fold enrichment criter- 3 .5. Hierarchical clustering analysis showed similaritiesion the number of ‘enriched’ genes decreased (Table 1). in expression patterns between the two divisions of theThe highest number of ‘enriched’ genes for each criterion PVN and between the three regions of the hippocampalwas found in LC. formation

3 .3. Known expression of ‘signature’ genes is consistent To compare the overall gene expression patterns of thewith cDNA microarray analysis different brain nucleus, hierarchical clustering was carried

out (Fig. 4). Three sets of data were used for clustering:Previous published reports corroborate many of our ‘all’ 2145 genes (Fig. 4a), 1402 ‘expressed’ genes (Fig.

findings (Table 2). For example, in our study, abundant 4b), and 254 ‘enriched’ genes (1.5-fold criteria) (Fig. 4c).mRNA expression encoding vasopressin (GB: AI072073) Similarities were observed in expression patterns betweenand oxytocin (GB: M88355) was detected in PA and MG the two divisions of the PVN in all three methods (Fig. 4).cells but not in the other examined nuclei. This is in The dendrograms show that the two divisions of the PVNagreement with previous in-situ hybridization (ISH) (PA and MG) were joined by a very short branch reflectingstudies looking at those same genes [3,7,24]. Examples of their similarities compared to the other nuclei. Using ‘all’enriched genes within one subnucleus were also observed. genes, the clustering does not compartmentalize CA1, CA3Somatostatin (GB: AA866347) mRNA expression within and DG (Fig. 4a). But when ‘expressed’ or ‘enriched’the PVN was confined to the PA division consistent with a genes are used to define the relatedness of the nuclei,previous report (ISH) [17]. Dynorphin mRNA was selec- similarities between the three regions of the hippocampaltively detected in DG but not in CA1 or CA3, matching the formation are seen (Fig. 4b,c). In Fig. 4b,c, the DR and LCdistribution reported by others via ISH [24]. Again within were located on a different branch of the dendrogram,

P.B

onaventureet

al./

Brain

Research

943(2002)

38–47

43

Table 2Distribution of several ‘signature genes’ for specific laser-captured brain nuclei. Data represent the median fluorescence intensity (695% confidence interval). Note that the fluorescence intensityreadings for vasopressin and oxytocin in MG correspond to the maximal intensity. The anatomical distribution of these genes matched the distribution previously described in the literature. Referencesare shown in the last column. The colors represent the fluorescence intensity, from yellow to red: yellow being weakly expressed and red highly expressed

44P.

Bonaventure

etal.

/B

rainR

esearch943

(2002)38

–47

Table 3List of top 5 ‘enriched’ genes (with a GeneBank accession number) found to be ‘enriched’ (see text for definition) in one of the nucleus /subnucleus investigated with respect to all six othernuclei / subnuclei. Data represent the median fluorescence intensity 6 95% confidence interval. Data are listed by decreasing ratio. L Reference available in the literature (see Table 2) P Verified byReal-Time PCR (see Fig. 3)

P. Bonaventure et al. / Brain Research 943 (2002) 38 –47 45

Fig. 3. Real-Time PCR. Four genes were arbitrarily selected for further data validation using Real-Time PCR. Cyclophilin (GB: AA818858) was chosen asa positive control for each assay, and used for normalization of both cDNA microarray and Real-Time PCR data. Genes with the highest value in eachsample were set to 100%, and other data points were all relative to that. Standard deviations were used to indicate the variability of the data. The open barsrepresent data from microarray experiments and the solid bars represent data from Real-Time PCR experiments.

3 .6. ‘Enriched’genes are putative nuclei specific markers

A list containing the top five ‘enriched’ genes (forwhich a GB accession number was available) in one of thenucleus or subnucleus relative to all six other is given inTable 3. The enrichment fold is also provided in Table 3.

3 .7. The presence of serotonin 5-HT receptor mRNA2B

in the investigated nuclei was confirmed by in-situhybridization

Fig. 4. Hierarchical cluster analysis. Hierarchical cluster analysis wasgenerated using Cluster (see Material and methods for details). The seven

A hybridization signal statistically different from theregions were assembled into a dendrogram, where items are joined byvery short branches if they are similar to each other and by increasingly background was observed in all the investigated nuclei forlonger branches as they become dissimilar. Different numbers of genes 5-HT receptor mRNA (median fluorescence2Bwere used for the analysis: (a) 2145 genes; (b) 1402 ‘expressed’ genes intensity695% confidence interval: CA1, 739.01643.28;(see text for definition); (c) 254 ‘enriched’ genes (see text for definition).

CA3, 189.32615.29; DG, 904.886122.14; DR,Note that in (b) and (c), the two divisions of the PVN (PA and MG) were408.47636.64; LC, 1013.016277.72; MG, 182.79611.61;joined by a very short branch, reflecting their similarities compared to the

35other areas. Similarly, in (c), the three regions of the hippocampal PA, 287.75619.17). ISH using S-labeled riboprobes wasformation (CA1, CA3 and DG) were also located on a common branch. carried out to confirm this observation. Incubation with the

sense probe did not yield a hybridization signal (Fig. 5a).Hybridization signals with the antisense probe are shownin Fig. 5b (hippocampal formation) and Fig. 5c (DR).

suggesting that these nuclei have greater differences in Hybridization signal was also detected in habenula (Fig.gene expression among the seven nuclei that were ex- 5b), cortex (Fig. 5b,c), LC, PVN, cerebellum and severalamined. brain stem nuclei (data not shown).

46 P. Bonaventure et al. / Brain Research 943 (2002) 38 –47

underlying the pair-wise similarities of gene expressionprofiles within one nucleus obtained in the present studydemonstrates the reproducibility of the LCM, T7-basedRNA amplification and cDNA microarray approach (Fig.2a). The expression profile of a set of four genes arbitrarilyselected was verified by Real-Time PCR. Similar geneexpression patterns were found with both techniques (Fig.3). However, as expected, higher sensitivity was observedwith Real-Time PCR. The high concordance between ourdata and published results for 17 genes further validatesthe approach we took (Table 2). Furthermore, our methodwas a more powerful one because we were able todemonstrate specific gene expression patterns within sub-nuclei that are closely related. For example, the expressionof NF-L (GB: MN031783) within the hippocampal forma-tion was enriched in CA3 whereas the expression of Trp6A(GB: AB051212) was enriched in DG. These results werefurther confirmed by Real-Time PCR. The preferentialexpression of Trp6A in dentate gyrus has been previouslyreported by ISH [15]. To the best of our knowledge, the

Fig. 5. Distribution of 5-HT receptor mRNA in the hippocampal2B preferential expression of the neurofilament subunit NF-Lformation and DR. Distribution of 5-HT receptor mRNA in pseudo-2B in CA3 has not been previously reported. Neurofilamentscolor pictures representing autoradiograms generated with the Fujifilm

have a well-established role in the control of axon caliberBio-Imaging Analyzer System of frontal sections through the rat brainand there is growing evidence that neurofilaments can(left hemisphere). Colors represent relative levels of optical density,

ranging from red.yellow.green.light blue.dark blue. (a) ISH spe- affect the dynamics and function of other cytoskeletal35cificity control, S-sense probe for the 5-HT receptor mRNA, note the2B elements such as microtubules and actin filaments [9].35absence of specific labeling. (b) S-antisense probe of 5-HT receptor2B Within the PVN, as previously reported [16], dynorphinmRNA in the hippocampal formation and (c) DR. Note in (b) the

(GB: M10088) was highly enriched in the MG division.hybridization signal in the habenula (Ha) and in cortical and other brainThese ‘enriched’ genes could be important in defining thestem regions (c). Scale bar represents 0.3 cm.

unique functions of these nuclei. Other examples of‘enriched’ genes are listed in Table 3.

4 . Discussion The presence of a low level of vasopressin mRNA in thehippocampal formation has been reported using ISH [7]. In

By integrating LCM, T7-based RNA amplification and the present study, we did not detect vasopressin mRNA incDNA microarrays, we have generated a catalogue of the hippocampal formation. This discrepancy most likelyputative cell-specific gene expression profiles that can be reflects differences in assay sensitivity between ISH andsubsequently used as nuclei specific markers. Expression cDNA microarray.analysis can provide insight into one potential source of We were able to classify the different laser-capturedfunctional differences between brain nuclei. Recently, brain nuclei by hierarchical clustering analysis. As ex-Sandberg et al. [20] and Zirlinger et al. [30] have shown pected, we found that the two divisions of the PVN (PAthe utility of brain region expression profiling on grossly and MG) were closely related, as were the three regions ofdissected brain regions, which represent multiple nuclei the hippocampal formation. Higher correlation coefficientsand heterogenous population of neurons and glia cells. We observed within subnuclei (PA/MG or CA1/CA3/DG)present here an approach that is neuroanatomically defined than between different nuclei (Fig. 2b) reinforces this(via LCM, Fig. 1) and allows the procurement of relatively point. These results are not surprising, because from apure neuronal populations. In addition, a single-dye label- developmental point of view, these cell populations (i.e.ing method was used for the present cDNA microarray MG and PA or CA1, CA3 and DG) are closely related. Thestudy. This allows the comparison of many samples, in dendrograms shown in Fig. 4 represent an approach tocontrast to dual color microarray experiments where all the classify brain nuclei based on expression profiles. Ourpossible pair-wise comparisons have to be competitively hierarchical cluster analysis demonstrates that each of thehybridized together. seven nuclei had a unique gene expression profile and that

In order to demonstrate the feasibility of gene expres- there is a molecular basis for the previously definedsion mapping studies in laser-captured rat brain nuclei, 100 anatomic nuclei or subnuclei.cells from the following brain nuclei were analyzed: LC, Interestingly, our microarray data shows the presence ofDR, PVN (PA and MG divisions) and hippocampal the serotonin 5-HT receptor mRNA in all the nuclei2B

formation (CA1, CA3 and DG). The correlation coefficient investigated. This finding was further confirmed by our

P. Bonaventure et al. / Brain Research 943 (2002) 38 –47 47

expression of the human histamine H3 receptor, Mol. Pharmacol. 55results from ISH. This is a novel finding since the(1999) 1101–1107.distribution of 5-HT receptor mRNA in these nuclei2B [13] L. Luo, R.C. Salunga, H. Guo, A. Bittner, K.C. Joy, J.E. Galindo, H.(hippocampal formation, PVN, DR and LC) has not beenXiao, K.E. Rogers, J.S. Wan, M.R. Jackson, M.G. Erlander, Gene

reported before. Evidence for expression of 5-HT re-2B expression profiles of laser-captured adjacent neuronal subtypes,ceptor protein has been shown in cerebellum, septum, Nat. Med. 5 (1999) 117–122.

[14] T.R. Madhav, Q. Pei, T.S. Zetterstrom, Serotonergic cells of the ratdorsal hypothalamus and amygdala by immunocytoch-raphe nuclei express mRNA of tyrosine kinase B (trkB), the high-emistry [4]. Of particular interest is the presence of 5-HT2Baffinity receptor for brain derived neurotrophic factor (BDNF),mRNA in DR, which suggests the presence of an addition-Brain Res. Mol. Brain Res. 93 (2001) 56–63.al serotonin autoreceptor in rat brain. Accounts pertaining

[15] N. Mizuno, S. Kitayama, Y. Saishin, S. Shimada, K. Morita, C.to the functional role of 5-HT receptor in the central2B Mitsuhata, H. Kurihara, T. Dohi, Molecular cloning and characteri-nervous system are limited and research has been ham- zation of rat trp homologues from brain, Brain Res. Mol. Brain Res.

64 (1999) 41–51.pered by the lack of selective pharmacological agents. Our[16] B.J. Morris, I. Haarmann, B. Kempter, V. Hollt, A. Herz, Localiza-findings suggest an interesting role for 5-HT receptor in2B tion of prodynorphin messenger rna in rat brain by in situ hybridiza-the brain. Hence, our study has successfully demonstrated

tion using a synthetic oligonucleotide probe, Neurosci. Lett. 69the importance, feasibility and utility of cellular brain (1986) 104–108.nuclei profiling. [17] J.V. Priestley, M. Rethelyi, P.K. Lund, Semi-quantitative analysis of

somatostatin mRNA distribution in the rat central nervous systemusing in situ hybridization, J. Chem. Neuroanat. 4 (1991) 131–153.

[18] J.M. Reul, E.R. de Kloet, Two receptor systems for corticosterone inA cknowledgementsrat brain: microdistribution and differential occupation, Endocrinolo-gy 117 (1985) 2505–2511.

We thank Dr S. Chaplan for her advice on Real-Time [19] R.C. Salunga, H. Guo, L. Luo, A. Bittner, K.C. Joy, J. Chambers, J.Wan, M.R. Jackson, M.G. Erlander, Gene expression analysis viaPCR.cDNA microarrays of laser capture microdissected cells from fixedtissue, in: M. Schena (Ed.), DNA Microarrays: A Practical Ap-proach, Oxford University Press, Oxford, 1999.

R eferences [20] R. Sandberg, R. Yasuda, D.G. Pankratz, T.A. Carter, J.A. Del Rio, L.Wodicka, M. Mayford, D.J. Lockhart, C. Barlow, Regional andstrain-specific gene expression mapping in the adult mouse brain,[1] P. Bonaventure, P. Voorn, W.H. Luyten, M. Jurzak, A. Schotte, J.E.Proc. Natl. Acad. Sci. USA 97 (2000) 11038–11043.Leysen, Detailed mapping of serotonin 5-HT and 5-HT receptor1B 1D

[21] D. Shalon, Gene expression micro-arrays: a new tool for genomicmessenger RNA and ligand binding sites in guinea-pig brain andresearch, Pathol. Biol. (Paris) 46 (1998) 107–109.trigeminal ganglion: clues for function, Neuroscience 82 (1998)

[22] R. Shigemoto, S. Nakanishi, N. Mizuno, Distribution of the mRNA469–484.for a metabotropic glutamate receptor (mGluR1) in the central[2] P.O. Brown, D. Botstein, Exploring the new world of the genomenervous system: an in situ hybridization study in adult and develop-with DNA microarrays, Nat. Genet. 21 (1999) 33–37.ing rat, J. Comp. Neurol. 322 (1992) 121–135.[3] J.P. Burbach, T.A. Voorhuis, H.H. van Tol, R. Ivell, In situ

[23] M.C. Swan, A.R. Najlerahim, J.P. Bennett, Expression of serotoninhybridization of oxytocin messenger RNA: macroscopic distributiontransporter mRNA in rat brain: presence in neuronal and non-and quantitation in rat hypothalamic cell groups, Biochem. Biophys.neuronal cells and effect of paroxetine, J. Chem. Neuroanat. 13Res. Commun. 145 (1987) 10–14.(1997) 71–76.[4] M.S. Duxon, T.P. Flanigan, A.C. Reavley, G.S. Baxter, T.P. Blac-

[24] M. Tohyama, K. Takatsuji, Atlas of Neuroactive Substances andkburn, K.C. Fone, Evidence for expression of the 5-hydroxy-their Receptors in the Rat, Oxford University Press, Oxford, 1998.tryptamine-2B receptor protein in the rat central nervous system,

[25] M.E. Trulson, M.S. Cannon, J.D. Raese, Identification of dopamine-Neuroscience 76 (1997) 323–329.containing cell bodies in the dorsal and median raphe nuclei of the[5] S. Fitzpatrick-McElligott, J.P. Card, M.E. Lewis, F. Baldino Jr.,rat brain using tyrosine hydroxylase immunochemistry, Brain Res.Neuronal localization of prosomatostatin mRNA in the rat brainBull. 15 (1985) 229–234.with in situ hybridization histochemistry, J. Comp. Neurol. 273

[26] S.J. Watson, F. Meng, R.C. Thompson, H. Akil, The ‘chip’ as a(1988) 558–572.specific genetic tool, Biol. Psychiatry 48 (2000) 1147–1156.[6] D.H. Geschwind, Mice, microarrays, and the genetic diversity of the

[27] S.M. Welford, J. Gregg, E. Chen, D. Garrison, P.H. Sorensen, C.T.brain, Proc. Natl. Acad. Sci. USA 97 (2000) 10676–10678.Denny, S.F. Nelson, Detection of differentially expressed genes in[7] M. Hallbeck, O. Hermanson, A. Blomqvist, Distribution of pre-primary tumor tissues using representational differences analysisprovasopressin mRNA in the rat central nervous system, J. Comp.coupled to microarray hybridization, Nucleic Acids Res. 26 (1998)Neurol. 411 (1999) 181–200.3059–3065.[8] S. Horvath, E. Mezey, M. Palkovits, Partial coexistence of growth

hormone-releasing hormone and tyrosine hydroxylase in paraven- [28] G.L. Yao, H. Kiyama, M. Tohyama, Distribution of GAP-43 (B50/tricular neurons in rats, Peptides 10 (1989) 791–795. F1) mRNA in the adult rat brain by in situ hybridization using an

[9] J.P. Julien, W.E. Mushynski, Neurofilaments in health and disease, alkaline phosphatase labeled probe, Brain Res. Mol. Brain Res. 18Prog. Nucleic Acid Res. Mol. Biol. 61 (1998) 1–23. (1993) 1–16.

[10] R.M. Lindsay, S.J. Wiegand, C.A. Altar, P.S. DiStefano, Neuro- [29] W.S. Young, M. Warden, E. Mezey, Tyrosine hydroxylase mRNA istrophic factors: from molecule to man, Trends Neurosci. 17 (1994) increased by hyperosmotic stimuli in the paraventricular and supra-182–190. optic nuclei, Neuroendocrinology 46 (1987) 439–444.

[11] D.J. Lockhart, C. Barlow, Expressing what’s on your mind: DNA [30] M. Zirlinger, G. Kreiman, D.J. Anderson, Amygdala-enriched genesarrays and the brain, Nat. Rev. Neurosci. 2 (2001) 63–68. identified by microarray technology are restricted to specific

[12] T.W. Lovenberg, B.L. Roland, S.J. Wilson, X. Jiang, J. Pyati, A. amygdaloid subnuclei, Proc. Natl. Acad. Sci. USA 98 (2001) 5270–Huvar, M.R. Jackson, M.G. Erlander, Cloning and functional 5275.